A control system consists of a controller, control elements, and controlled elements. In a typical velocity servo system, the control elements can be power transistors or SCR's, the controlled elements are usually an electric motor coupled to a load, and the controller compares a velocity command to a velocity feedback to derive a velocity error signal used to drive the power transistors or SCR's. Both dynamic and static parameters are associated with each section of the control system and these parameters must be properly matched to obtain optimum performance and to avoid instability problems. In most systems, the parameters of the controller are set to "compensate" for parameters of the controlled elements.
The conventional controller configuration found in most velocity servo systems is PID (proportion, integral with derivative-feedback). Compensation is achieved by adjusting four components which are the integrating and differentiating capacitors and their respective series resistors. Unfortunately, the four adjustable parameters are interrelated and are normally set for particular load condition though a trial and error procedure. In cases where load conditions are fixed, the optimum compensation can be fixed. On the other hand, where the load or plant conditions vary, such as in a robotics system, the compensation must be set for either the worst case conditions, or at a compromise setting. Pre-set compensations cannot achieve optimum performance for all load variations, and, hence, the system operates at below optimum performance at least part of the time.
Adaptive control concepts are known, but have not previously been practical. In adaptive systems, the controller senses the load conditions and automatically adjusts the controller compensation accordingly. Although the concept is simple, the implementation has proven to be impractical. To adjust multiple interrelated controller parameters according to dynamically varying load parameters requires complex modeling of system characteristics and an inordinate number of iterative calculations. Even relatively simple control systems in the past have required large size dedicated computers if there was to be any realistic hope of achieving real time adaptive control.